量子為主流企業(yè)應(yīng)用準(zhǔn)備就緒

時(shí)間：2022-04-18 03:42:01 | 來(lái)源：行業(yè)動(dòng)態(tài)

時(shí)間：2022-04-18 03:42:01 來(lái)源：行業(yè)動(dòng)態(tài)

While the technology has yet to break cryptography, quantum computing is ready for mainstream adoption and already is tapped to address real-world enterprise challenges.

雖然量子計(jì)算技術(shù)還不能破解密碼系統(tǒng)，但該技術(shù)已經(jīng)準(zhǔn)備好了進(jìn)入主流應(yīng)用，也確實(shí)已經(jīng)開(kāi)始用于解決現(xiàn)實(shí)世界的企業(yè)挑戰(zhàn)。

Pointing specifically to D-Wave's proprietary annealing technology, Baratz said this allowed quantum computing to scale more easily and be less sensitive to noise and computational errors, to which gate-based systems were prone. Currently in its fifth generation, D-Wave's quantum computers clock more than 5,000 qubits and capable of supporting commercial rollout "at commercial scale", he said.

Baratz特別提到D-Wave的專有退火技術(shù)，他表示，D-Wave的專有退火技術(shù)使得量子計(jì)算更容易擴(kuò)展，該技術(shù)對(duì)噪聲和計(jì)算錯(cuò)誤的敏感度更低，而基于門(mén)的系統(tǒng)就容易出現(xiàn)這方面的問(wèn)題。他表示，目前D-Wave的量子計(jì)算機(jī)已發(fā)展到了第五代，算力達(dá)5,000量子比特，能夠支持 "商業(yè)規(guī)模 "的商業(yè)推廣。

This, he added, was a stage that no other market players had been able to achieve thus far with the gate-based model. Commonly adopted in the industry today, the gate system made quantum computers tough to build and sensitive error. Its most stable state currently generated about 30 qubits, which was sufficient to power mostly research work and unlikely to be used to solve business problems at scale for another seven to 10 years, he said.

他補(bǔ)充表示，迄今為止，D-Wave第五代的算力是其他市場(chǎng)參與者都沒(méi)有能夠用基于門(mén)的模式實(shí)現(xiàn)的。目前業(yè)界普遍采用的門(mén)系統(tǒng)用于制造量子計(jì)算機(jī)的難度很大，而且對(duì)錯(cuò)誤敏感。他表示，目前這一類量子計(jì)算機(jī)最穩(wěn)定的狀態(tài)大約能達(dá)到30個(gè)量子比特，用于大部分研究工作夠了，但在未來(lái)7到10年內(nèi)不太可能大規(guī)模用于解決商業(yè)問(wèn)題。

"Error rates on [gate-based systems] are so high you can't really do anything with them, even with small problems," he added, noting that a competitor last year said it was able to solve a specific optimisation problem on its quantum computer. However, this was possible once out of every 100,000 attempts, he said.

他還表示，"基于門(mén)的量子系統(tǒng)的錯(cuò)誤率非常高，沒(méi)法真正用來(lái)做任何事情，即使用于解決小問(wèn)題都不行。"他指出，一個(gè)競(jìng)爭(zhēng)對(duì)手去年曾表示能夠在自己的量子計(jì)算機(jī)上解決一個(gè)特定的優(yōu)化問(wèn)題。但他表示，要嘗試了10萬(wàn)次才可能。

Quantum computing runs on principles of quantum mechanics that include probabilistic computation.

量子計(jì)算基于量子力學(xué)原理運(yùn)作，其中包括概率計(jì)算。

Baratz said annealing technology, designed specifically for optimisation purposes, had a higher influence on the probability of outcomes and, hence, was less sensitive to errors. It also learnt from where it ended with the previous computation to finetune future ones.

Baratz表示，專門(mén)為優(yōu)化目的而設(shè)計(jì)的退火技術(shù)對(duì)結(jié)果概率的影響較大，因此對(duì)錯(cuò)誤的敏感度較低。退火技術(shù)還能從過(guò)去的計(jì)算結(jié)束的地方開(kāi)始學(xué)習(xí)，可以微調(diào)未來(lái)的計(jì)算。

"When you lose coherence, you end up with garbage. With annealing, when you lose coherence, you settle into a [potential] solution and restart the computation to try and improve the solution," he said. Gate-based model, in comparison, could not do that since it would lose coherence after every computation rather than pick off from the previous run.

他表示，"失去了連貫性時(shí)，最終就會(huì)得到垃圾。退火技術(shù)可以在失去連貫性時(shí)，停在一個(gè)潛在的解決方案里并可以重新開(kāi)始計(jì)算以及改進(jìn)解決方案。"而相比之下，基于門(mén)的模型做不到這一點(diǎn)，因?yàn)檫@種模型在每次計(jì)算后都失去了一致性，沒(méi)法接著之前的運(yùn)算繼續(xù)下去。

A grocery using D-Wave to enhance a portion of the customer's logistics system was able to solve an optimisation problem in two minutes per week per location, where previously it took 25 hours per week per location, he noted.

他指出，一家雜貨店可以用D-Wave加強(qiáng)客戶的部分物流系統(tǒng)，解決每個(gè)地點(diǎn)每周的優(yōu)化問(wèn)題只需兩分鐘，而以前解決每個(gè)地點(diǎn)每周的優(yōu)化問(wèn)題需要25小時(shí)。

There currently are more than 20,000 developers worldwide that have signed up to access Leap, with some 1,000 regularly using the service each month. Paying customers fork out an estimated ,000 an hour to run computations on D-Wave computers.

目前，全球已有超過(guò)2萬(wàn)名開(kāi)發(fā)者注冊(cè)使用Leap，每月約有1,000人定期使用該項(xiàng)服務(wù)。付費(fèi)客戶在D-Wave計(jì)算機(jī)上運(yùn)行計(jì)算估計(jì)每小時(shí)需支付2,000美元。

Baratz noted, though, that its systems could not solve all quantum computing issues because annealing was designed specifically to solve optimisation problems, which were common challenges for businesses. Gate-based systems, on the other hand, would be able to solve any computation problems once the error rates were reduced -- something he said likely would not actualise for at least another seven years.

不過(guò)Baratz指出，D-Wave系統(tǒng)也無(wú)法解決所有量子計(jì)算問(wèn)題，因?yàn)橥嘶鸺夹g(shù)是專門(mén)為解決優(yōu)化問(wèn)題而設(shè)計(jì)的，優(yōu)化問(wèn)題是企業(yè)面臨的共同挑戰(zhàn)。而基于門(mén)的系統(tǒng)在錯(cuò)誤率降低后將能夠解決任何計(jì)算問(wèn)題，但他表示，降低錯(cuò)誤率很可能至少還需要7年才能實(shí)現(xiàn)。

So while D-Wave's annealing-powered quantum computers were limited to solving optimisation problems, they were capable of addressing real-world business challenges today, he said. Its systems also were on a path towards building a universal error correction system by leveraging the technology it had, he added.

他表示，因此，雖然D-Wave基于退火技術(shù)的量子計(jì)算機(jī)僅限于解決優(yōu)化問(wèn)題，但這種量子計(jì)算機(jī)有能力解決當(dāng)今現(xiàn)實(shí)世界的商業(yè)挑戰(zhàn)。他補(bǔ)充表示，D-Wave系統(tǒng)也可以利用所擁有的技術(shù)走構(gòu)建通用糾錯(cuò)系統(tǒng)之路。

To date, more than 250 applications had been built with D-Wave systems, most of which used Leap and spanned various use cases including financial modelling, scheduling, protein folding, and manufacturing optimisation, the vendor said.

D-Wave表示，到目前為止，已經(jīng)有超過(guò)250個(gè)應(yīng)用的構(gòu)建用上了D-Wave系統(tǒng)，其中大部分使用Leap，應(yīng)用涵蓋了各種用例，包括財(cái)務(wù)建模、調(diào)度、蛋白質(zhì)折疊和制造優(yōu)化等等。